Intelligent System for Predicting Bank Policy Acceptance by Ensemble Machine Learning and Model Explanation

Efficient management of financial resources is crucial for the sustainability and competitiveness of banks, particularly in optimizing term deposit subscriptions to maintain liquidity. This paper introduces an advanced intelligent system for predicting term deposit acceptance using ensemble machine learning techniques. Our approach combines Random Forest and K-Nearest Neighbors (KNN) models to enhance prediction accuracy while providing clear explanations. The system follows the CRISP-DM methodology, which includes detailed phases of data preparation, modeling, fine-tuning, and model explanation. We utilize Random Forest for its feature importance metrics and KNN for assessing feature relevance through nearest neighbor analysis. The integration of these methods allows us to generate comprehensive explanations of prediction outcomes by identifying and interpreting key features influencing decision-making. By applying this method to the Bank Marketing Data Set, we demonstrate improved performance across standard metrics such as accuracy, precision, recall, and F1-score. The detailed explanation phase helps understand the model’s decision process, providing actionable insights for refining telemarketing strategies. This research presents a robust framework for implementing explainable machine learning in financial marketing, enhancing both predictive accuracy and interpretability for better-informed decision-making.